Topic modeling and Sentiment Analysis-based Recommender system: A literature review

Recommender Systems (RSs) are widely used to support decision-making process for users by providing recommendations to help them choose the suited products or services based on their history preferences and profiles. The difficulty of obtaining personal user history has brought into attention the need to refer to another alternative when designing a RS and not simply relying on users' ratings. Over the past decade, various websites provided a sharing space for users to freely give reviews and textual recommendations on various commodities. The reviews may contain rich information related to the item or a specific feature that can be extracted using natural language processing techniques. This survey provides an overview of RSs as well as possible techniques that could be deployed on user reviews to enhance the recommendation accuracy. We mainly focus on two techniques: Topic modeling which extracts main features or topics of a given document and Sentiment analysis which studies user polarity through their associated textual input. We show the advantages of associating these approaches with RSs and discuss their main applications, challenges, and open research directions.